I designed and built a board with multiple Transputers on. We didn't have the Occam development software so I had to write an assembler, which I did in BBC Basic as the transputer board interface connected to the BBC micro.
Transputer instructions are all variable length, and their length depends on their arguments. This includes jump instructions which used relative addressing. This meant that the assembler would take as many as 7 passes before the output converged into the optimum result.
Because the BBC micro had so little memory (32k) the intermediate result of each pass had to be written to floppy disk! I seem to remember it taking about 15mins with an awful lot of clonk clonk clonk from the floppy disk to get my code assembled.
I finally got my physics simulations (Ising spin models) assembled and with 3 Transputers they ran a little bit faster than the Fortran program I'd written for the university mainframe which I was very pleased about.
I designed and built a board with multiple Transputers on. We didn't have the Occam development software so I had to write an assembler, which I did in BBC Basic as the transputer board interface connected to the BBC micro.
Transputer instructions are all variable length, and their length depends on their arguments. This includes jump instructions which used relative addressing. This meant that the assembler would take as many as 7 passes before the output converged into the optimum result.
Because the BBC micro had so little memory (32k) the intermediate result of each pass had to be written to floppy disk! I seem to remember it taking about 15mins with an awful lot of clonk clonk clonk from the floppy disk to get my code assembled.
I finally got my physics simulations (Ising spin models) assembled and with 3 Transputers they ran a little bit faster than the Fortran program I'd written for the university mainframe which I was very pleased about.
Happy days!